Clinical AI Harm, Smarter Reasoning, and Safer Agents
Three papers: AI safety measures withhold critical clinical guidance from patients, SAT cuts reasoning tokens by 40%, and conformal prediction blocks wrong multi-agent consensus.
Articles Tagged "Benchmarks"
Muse Spark Review: Strong on Health, Weak on Code
Meta's first proprietary frontier model leads on HealthBench Hard and scientific reasoning but trails rivals in coding and agentic tasks - with no public API yet.
Gemini 2.5 Flash vs Claude Sonnet 4.6: Cost vs Code
Gemini 2.5 Flash costs 10x less and runs 4x faster than Claude Sonnet 4.6, but trails badly on coding benchmarks - here is the full breakdown.
Instruction Following Leaderboard: IFEval Rankings 2026
Rankings of AI models on IFEval and IFBench, the two main benchmarks for measuring how reliably LLMs follow precise formatting, length, and content constraints.
EXAONE 4.5: LG's Open VLM Beats GPT-5-mini on STEM
LG AI Research released EXAONE 4.5, a 33B open-weight vision-language model that posts higher STEM scores than GPT-5-mini and Claude 4.5 Sonnet - but a non-commercial license caps its real-world reach.
Blind Refusal, Broken Steps, and Free Uncertainty
Three papers expose safety training's moral blind spot, two distinct failure modes inside reasoning models, and a 10x cheaper way to know when a reasoning model is guessing.
Muse Spark
Meta's first closed-source frontier model scores 52 on the Artificial Analysis Intelligence Index, leads on HealthBench Hard, and ships free at meta.ai - but has no public API yet.
Best AI Models for Agentic Tool Use - April 2026
Claude Opus 4.6 leads SWE-bench Verified at 80.8% and OSWorld at 72.7% for agentic tasks, while GPT-5.4 ties for computer use; no single model dominates every workflow type.
Meta Muse Spark Launches, Ranks 4th Among Frontier Models
Meta Superintelligence Labs releases Muse Spark, its first AI model built from scratch in nine months, landing 4th on the Artificial Analysis Intelligence Index.
GLM-5.1 Tops SWE-Bench Pro With Zero NVIDIA Hardware
Z.ai's GLM-5.1 scores 58.4 on SWE-bench Pro, edging out GPT-5.4 and Claude Opus 4.6, after being trained on 100,000 Huawei Ascend chips with no US silicon.
Claude Mythos Preview Finds Thousands of Zero-Days
Anthropic's restricted Claude Mythos Preview model autonomously discovered thousands of high-severity vulnerabilities across every major OS and browser, including bugs hiding in plain sight for 27 years.
AutoAgent Builds Its Own Harness, Tops Two Benchmarks
Kevin Gu's MIT-licensed AutoAgent lets a meta-agent engineer and hill-climb its own agent harness overnight, claiming the top GPT-5 slot on TerminalBench and first place on SpreadsheetBench.
